Skip to main content

Help - Unexpected results

Sometimes you get unexpected results. This chapter should help you to understand why this happens and how to solve it. Generally as a first step it is helpful to check the protocol / logs of the test case

Unexpected steps created​

In case you see steps created that you did not expect, this can have multiple reasons. Most likely you are trying to create actions which are not supported.

Please try the following steps:

  1. Check the prompting technique in the Test-Case Definition whether you can find a better way to describe the action.
  2. Check the "Explain" button in the Test-Case Definition for an explanation of how your action was interpreted.
  3. Write the step in english and check if the AI agent understands it correctly.
  4. Avoid to refer to other steps / previous results in your description. The AI agent does not have a memory of previous steps (only via parameters).

Unexpected elements are identified for actions (Click, Type, ...)​

When trying to interact with elements on the page, sometimes the AI agent might identify elements you were not expecting. You can use the following tips to solve the issue.

  1. Is your element description actually unique? In case of a vague description, the AI element might take one element in one run and another element in another run. Try to be more specific by e.g. providing the description of a section or parent element.
  2. Try to activate enhanced reasoning for the step ( see Advanced Settings in the step details for more details).
  3. Check the step in Recording Mode to change the element identified.
  4. Discuss the step in Recording Mode with the AI agent to understand why the AI agent identified the element.
  5. Check if the element is active for the interaction. The AI agent can only interact with active elements (not disabled, ...) - depending on your application settings.
  6. Try to rephrase the step as per the Prompting Technique to get a better result.
  7. If the wrong element is identified, but the provided xpath seems to be ok, ask the developer of the application to implement better IDs/Classes for the element, so that it is possible to identify the element better.
  8. Ask a developer to give you a xpath for the element you want to interact with. You can use this xpath in the Test-Case Definition to interact with the element.
  9. Ask a developer to write a scripting function to identify the element you want to interact with. You can use this scripting function in the Test-Case Definition to interact with the element.

No elements are identified for actions (Click, Type, ...)​

When trying to interact with elements on the page, sometimes the AI agent might not identify any elements. You can use the following tips to solve the issue.

  1. Check if the required element is in an iframe. IFrames are supported, but must be explicitly activated for the page under test / tenant, as it makes test performance a little bit slower. Please activate the iframe processing in the Application Settings (Additional Functions) or in the tenant settings.
  2. Check if the required element is not visible. The AI agent can only interact with visible elements.
  3. Check if the required element is not loaded yet. The AI agent can only interact with elements that are loaded.
  4. Rephrase what you are searching based on the Prompting Technique.
  5. Define the element via xpath
  6. Define a scripting function to identify the element.

The assertion provided unexpected results​

In case the assertion provided unexpected results, you can use the following tips to solve the issue.

  1. Check the Prompting Technique to get a better result.
  2. Go into Recording Mode and discuss the step with the AI agent.
  3. Check the screenshots which were actually used by the AI agent in the execution details of the step. Especially in case of scrolling you might notice, that the screenshot does not show all elements.
  4. Are you asserting mathematical calculations or ask the AI agent to count? Counting and mathematical operations are hard to grasp for the AI agent. Try to avoid this.